Scribd
Upload
9 views13 pages

Prompt Engineering Process Using Key Elements

Uploaded by

Jagriti Bhatia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
9 views13 pages

Prompt Engineering Process Using Key Elements

Uploaded by

Jagriti Bhatia
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

Prompt Engineering

Important Elements
● Task Mandatory More Importan
● Context
● Exemplar Important
● Persona
● Format
● Tone Nice to have
Optional
Reason for higher to lower Importance
● If we give task alone to GPT it will still give something useful as an output but
if dont give task and only input Context, the output will be meaningless to us.
Note
Our Prompt will be combination of blocks mentioned in previous slide. It’s not
necessary that we need every block of element in our prompt.
#1 Task
● Always start the Task sentence with an Action Verb.
● Eg - Generate, Give, Analyze, Summarize, Write etc, for us, these can be -
analyze the product information, categorize the Product to a particular
category, generate confidence for each category
Using GPT to finalize the final Task
● Conversation Link -
https://chat.openai.com/share/2b59a21e-4d70-45a2-b8e9-ffd160a79117

● With Format : Analyze the provided product information to precisely


Categorize the product into one of the predefined categories. Additionally,
Generate a table with headers - 'Category Name' and 'Confidence Level'
(High, Medium, Low) for each assigned category. Ensure that there is no
extra information beyond the presented table.
#2 Context
● Limit the Endless Possibilities
● 3 Main Questions to ask before setting context -
○ What is the background of the task ? Eg - for us, it is that we are an ecommerce platform with
products and multiple categories and every product must be mapped to a single category
○ What does the success look like ? Eg - for us,most relevant and most specific category
○ What is the environment we are in ? Eg - for us, the rules for selection like using product
name primarily and then optionally using product desc, product isq, and then what if there is
no extra information or what if desc and name point to different things.
#3 Exemplars
● Basically using a relevant example or framework on how GPT must arrive at
the conclusion.
#2 Exemplars
● Following are some examples of exemplars
#3 Exemplars
● For us, Examples can include -
○ An example where using Product Description a relevant specific category was predicted
○ An example of an existing case, where in gpt used to give irrelevant result, but us explaining
how the correct answer to it would be the answer
○ Thinking framework, for eg - a case where only product name was enough to generate high
confidence answer , a case alone product description or isq were enough, a case where seller
context helped, a case of multiple different directions and how to choose the best one out of
all
○ Note - Exemplars are not necessary for every prompt but adding an example can improve
quality of prompt greatly
#4 Persona
● Who do you want the AI to be ?
○ Tip - Think of someone who has instant access to the task we are facing
○ In our case, it can be a Category Agent Specialist, ecommerce category manager at
Indiamart, Category Manager at Alibana/Walmart etc
○ These can be exact people too, if they are famous, for example Albert Einstein, Steve Jobs
etc
#5 Format
● Exactly visualize what the End Result should look like
● Examples of Formats - Table, Code Blocks, Dictionary, bullets, List,
paragraph, etc
● For eg, for our Use case it can be Table with columns such as Category
Name, Confidence, Rank…or a json structure as we are following currently
along with a 10 word Explanation.
#6 Tone
● Tone of Voice
● Feeling / Adjective of answer
● For us it can be - Clear, Logical, Intelligent, Smart, Perfectionist
● Examples of tone -
Example - Using all together

You might also like